From: Vitalie Vrabie on
assuming that the user hasn't tweaked the MTU size in any way, how
does the IP protocol determine the maximum size of a packet it can
send down the line?
From: Vitalie Vrabie on
in my NDIS IM i correct the values when OID_GEN_MAXIMUM_FRAME_SIZE and
OID_GEN_MAXIMUM_TOTAL_SIZE are returned to the upper level, but IP
doesn't seem to query them at all!
From: Maxim S. Shatskih on
> does the IP protocol determine the maximum size of a packet

For local destinations achievable without a gateway, IP asks the MAC drivers below.

For any destination which requires a gateway, IP uses the magic number of 576 (IIRC), which is the mandatory packet size to be supported by any IP network.

TCP is more complex - see the RFCs on "path MTU discovery".

--
Maxim S. Shatskih
Windows DDK MVP
maxim(a)storagecraft.com
http://www.storagecraft.com

From: Vitalie Vrabie on
On Apr 15, 7:59 pm, "Maxim S. Shatskih"
<ma...(a)storagecraft.com.no.spam> wrote:
> > does the IP protocol determine the maximum size of a packet
>
> For local destinations achievable without a gateway, IP asks the MAC drivers below.

as far as i've investigated, it does NOT. :(
From: Vitalie Vrabie on
On Apr 16, 3:43 am, Vitalie Vrabie <vitalie.vra...(a)gmail.com> wrote:
> On Apr 15, 7:59 pm, "Maxim S. Shatskih" <ma...(a)storagecraft.com.no.spam> wrote:
> > For local destinations achievable without a gateway, IP asks the MAC drivers below.
>
> as far as i've investigated, it does NOT. :(

let me be precise.
on win7 it DOES as Maxim said.
("netsh interface ipv4 show subinterface" has confirmed that: it shows
MTU=1478 which is 1500 minus those 22 bytes my IM chops out).

on xp - this command doesn't work and i didn't find any equivalent.
but from what i've found experimentally - it doesn't seem to query the
miniport below.